2 * Copyright (c) 2007-2012 Nicira, Inc.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public License
14 * along with this program; if not, write to the Free Software
15 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
19 #include <linux/etherdevice.h>
21 #include <linux/if_vlan.h>
22 #include <linux/jhash.h>
23 #include <linux/kconfig.h>
24 #include <linux/kernel.h>
25 #include <linux/list.h>
26 #include <linux/mutex.h>
27 #include <linux/percpu.h>
28 #include <linux/rcupdate.h>
29 #include <linux/rtnetlink.h>
30 #include <linux/compat.h>
31 #include <linux/version.h>
32 #include <net/net_namespace.h>
37 #include "vport-internal_dev.h"
39 static void ovs_vport_record_error(struct vport *,
40 enum vport_err_type err_type);
42 /* List of statically compiled vport implementations. Don't forget to also
43 * add yours to the list at the bottom of vport.h. */
44 static const struct vport_ops *vport_ops_list[] = {
45 &ovs_netdev_vport_ops,
46 &ovs_internal_vport_ops,
47 &ovs_geneve_vport_ops,
48 #if IS_ENABLED(CONFIG_NET_IPGRE_DEMUX)
56 /* Protected by RCU read lock for reading, ovs_mutex for writing. */
57 static struct hlist_head *dev_table;
58 #define VPORT_HASH_BUCKETS 1024
61 * ovs_vport_init - initialize vport subsystem
63 * Called at module load time to initialize the vport subsystem.
65 int ovs_vport_init(void)
67 dev_table = kzalloc(VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
76 * ovs_vport_exit - shutdown vport subsystem
78 * Called at module exit time to shutdown the vport subsystem.
80 void ovs_vport_exit(void)
85 static struct hlist_head *hash_bucket(struct net *net, const char *name)
87 unsigned int hash = jhash(name, strlen(name), (unsigned long) net);
88 return &dev_table[hash & (VPORT_HASH_BUCKETS - 1)];
92 * ovs_vport_locate - find a port that has already been created
94 * @name: name of port to find
96 * Must be called with ovs or RCU read lock.
98 struct vport *ovs_vport_locate(struct net *net, const char *name)
100 struct hlist_head *bucket = hash_bucket(net, name);
103 hlist_for_each_entry_rcu(vport, bucket, hash_node)
104 if (!strcmp(name, vport->ops->get_name(vport)) &&
105 net_eq(ovs_dp_get_net(vport->dp), net))
112 * ovs_vport_alloc - allocate and initialize new vport
114 * @priv_size: Size of private data area to allocate.
115 * @ops: vport device ops
117 * Allocate and initialize a new vport defined by @ops. The vport will contain
118 * a private data area of size @priv_size that can be accessed using
119 * vport_priv(). vports that are no longer needed should be released with
122 struct vport *ovs_vport_alloc(int priv_size, const struct vport_ops *ops,
123 const struct vport_parms *parms)
128 alloc_size = sizeof(struct vport);
130 alloc_size = ALIGN(alloc_size, VPORT_ALIGN);
131 alloc_size += priv_size;
134 vport = kzalloc(alloc_size, GFP_KERNEL);
136 return ERR_PTR(-ENOMEM);
138 vport->dp = parms->dp;
139 vport->port_no = parms->port_no;
141 INIT_HLIST_NODE(&vport->dp_hash_node);
143 if (ovs_vport_set_upcall_portids(vport, parms->upcall_portids)) {
145 return ERR_PTR(-EINVAL);
148 vport->percpu_stats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
149 if (!vport->percpu_stats) {
151 return ERR_PTR(-ENOMEM);
154 spin_lock_init(&vport->stats_lock);
160 * ovs_vport_free - uninitialize and free vport
162 * @vport: vport to free
164 * Frees a vport allocated with ovs_vport_alloc() when it is no longer needed.
166 * The caller must ensure that an RCU grace period has passed since the last
167 * time @vport was in a datapath.
169 void ovs_vport_free(struct vport *vport)
171 kfree((struct vport_portids __force *)vport->upcall_portids);
172 free_percpu(vport->percpu_stats);
177 * ovs_vport_add - add vport device (for kernel callers)
179 * @parms: Information about new vport.
181 * Creates a new vport with the specified configuration (which is dependent on
182 * device type). ovs_mutex must be held.
184 struct vport *ovs_vport_add(const struct vport_parms *parms)
190 for (i = 0; i < ARRAY_SIZE(vport_ops_list); i++) {
191 if (vport_ops_list[i]->type == parms->type) {
192 struct hlist_head *bucket;
194 vport = vport_ops_list[i]->create(parms);
196 err = PTR_ERR(vport);
200 bucket = hash_bucket(ovs_dp_get_net(vport->dp),
201 vport->ops->get_name(vport));
202 hlist_add_head_rcu(&vport->hash_node, bucket);
214 * ovs_vport_set_options - modify existing vport device (for kernel callers)
216 * @vport: vport to modify.
217 * @options: New configuration.
219 * Modifies an existing device with the specified configuration (which is
220 * dependent on device type). ovs_mutex must be held.
222 int ovs_vport_set_options(struct vport *vport, struct nlattr *options)
224 if (!vport->ops->set_options)
226 return vport->ops->set_options(vport, options);
230 * ovs_vport_del - delete existing vport device
232 * @vport: vport to delete.
234 * Detaches @vport from its datapath and destroys it. It is possible to fail
235 * for reasons such as lack of memory. ovs_mutex must be held.
237 void ovs_vport_del(struct vport *vport)
241 hlist_del_rcu(&vport->hash_node);
242 vport->ops->destroy(vport);
246 * ovs_vport_set_stats - sets offset device stats
248 * @vport: vport on which to set stats
249 * @stats: stats to set
251 * Provides a set of transmit, receive, and error stats to be added as an
252 * offset to the collected data when stats are retrieved. Some devices may not
253 * support setting the stats, in which case the result will always be
256 * Must be called with ovs_mutex.
258 void ovs_vport_set_stats(struct vport *vport, struct ovs_vport_stats *stats)
260 spin_lock_bh(&vport->stats_lock);
261 vport->offset_stats = *stats;
262 spin_unlock_bh(&vport->stats_lock);
266 * ovs_vport_get_stats - retrieve device stats
268 * @vport: vport from which to retrieve the stats
269 * @stats: location to store stats
271 * Retrieves transmit, receive, and error stats for the given device.
273 * Must be called with ovs_mutex or rcu_read_lock.
275 void ovs_vport_get_stats(struct vport *vport, struct ovs_vport_stats *stats)
279 /* We potentially have 3 sources of stats that need to be
280 * combined: those we have collected (split into err_stats and
281 * percpu_stats), offset_stats from set_stats(), and device
282 * error stats from netdev->get_stats() (for errors that happen
283 * downstream and therefore aren't reported through our
284 * vport_record_error() function).
285 * Stats from first two sources are merged and reported by ovs over
286 * OVS_VPORT_ATTR_STATS.
287 * netdev-stats can be directly read over netlink-ioctl.
290 spin_lock_bh(&vport->stats_lock);
292 *stats = vport->offset_stats;
294 stats->rx_errors += vport->err_stats.rx_errors;
295 stats->tx_errors += vport->err_stats.tx_errors;
296 stats->tx_dropped += vport->err_stats.tx_dropped;
297 stats->rx_dropped += vport->err_stats.rx_dropped;
299 spin_unlock_bh(&vport->stats_lock);
301 for_each_possible_cpu(i) {
302 const struct pcpu_sw_netstats *percpu_stats;
303 struct pcpu_sw_netstats local_stats;
306 percpu_stats = per_cpu_ptr(vport->percpu_stats, i);
309 start = u64_stats_fetch_begin_irq(&percpu_stats->syncp);
310 local_stats = *percpu_stats;
311 } while (u64_stats_fetch_retry_irq(&percpu_stats->syncp, start));
313 stats->rx_bytes += local_stats.rx_bytes;
314 stats->rx_packets += local_stats.rx_packets;
315 stats->tx_bytes += local_stats.tx_bytes;
316 stats->tx_packets += local_stats.tx_packets;
321 * ovs_vport_get_options - retrieve device options
323 * @vport: vport from which to retrieve the options.
324 * @skb: sk_buff where options should be appended.
326 * Retrieves the configuration of the given device, appending an
327 * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
328 * vport-specific attributes to @skb.
330 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
331 * negative error code if a real error occurred. If an error occurs, @skb is
334 * Must be called with ovs_mutex or rcu_read_lock.
336 int ovs_vport_get_options(const struct vport *vport, struct sk_buff *skb)
341 if (!vport->ops->get_options)
344 nla = nla_nest_start(skb, OVS_VPORT_ATTR_OPTIONS);
348 err = vport->ops->get_options(vport, skb);
350 nla_nest_cancel(skb, nla);
354 nla_nest_end(skb, nla);
358 static void vport_portids_destroy_rcu_cb(struct rcu_head *rcu)
360 struct vport_portids *ids = container_of(rcu, struct vport_portids,
367 * ovs_vport_set_upcall_portids - set upcall portids of @vport.
369 * @vport: vport to modify.
370 * @ids: new configuration, an array of port ids.
372 * Sets the vport's upcall_portids to @ids.
374 * Returns 0 if successful, -EINVAL if @ids is zero length or cannot be parsed
375 * as an array of U32.
377 * Must be called with ovs_mutex.
379 int ovs_vport_set_upcall_portids(struct vport *vport, struct nlattr *ids)
381 struct vport_portids *old, *vport_portids;
383 if (!nla_len(ids) || nla_len(ids) % sizeof(u32))
386 old = ovsl_dereference(vport->upcall_portids);
388 vport_portids = kmalloc(sizeof *vport_portids + nla_len(ids),
393 vport_portids->n_ids = nla_len(ids) / sizeof(u32);
394 vport_portids->rn_ids = reciprocal_value(vport_portids->n_ids);
395 nla_memcpy(vport_portids->ids, ids, nla_len(ids));
397 rcu_assign_pointer(vport->upcall_portids, vport_portids);
400 call_rcu(&old->rcu, vport_portids_destroy_rcu_cb);
406 * ovs_vport_get_upcall_portids - get the upcall_portids of @vport.
408 * @vport: vport from which to retrieve the portids.
409 * @skb: sk_buff where portids should be appended.
411 * Retrieves the configuration of the given vport, appending the
412 * %OVS_VPORT_ATTR_UPCALL_PID attribute which is the array of upcall
415 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room.
416 * If an error occurs, @skb is left unmodified. Must be called with
417 * ovs_mutex or rcu_read_lock.
419 int ovs_vport_get_upcall_portids(const struct vport *vport,
422 struct vport_portids *ids;
424 ids = rcu_dereference_ovsl(vport->upcall_portids);
426 if (vport->dp->user_features & OVS_DP_F_VPORT_PIDS)
427 return nla_put(skb, OVS_VPORT_ATTR_UPCALL_PID,
428 ids->n_ids * sizeof(u32), (void *) ids->ids);
430 return nla_put_u32(skb, OVS_VPORT_ATTR_UPCALL_PID, ids->ids[0]);
434 * ovs_vport_find_upcall_portid - find the upcall portid to send upcall.
436 * @vport: vport from which the missed packet is received.
437 * @skb: skb that the missed packet was received.
439 * Uses the skb_get_hash() to select the upcall portid to send the
442 * Returns the portid of the target socket. Must be called with rcu_read_lock.
444 u32 ovs_vport_find_upcall_portid(const struct vport *p, struct sk_buff *skb)
446 struct vport_portids *ids;
449 ids = rcu_dereference(p->upcall_portids);
451 if (ids->n_ids == 1 && ids->ids[0] == 0)
454 hash = skb_get_hash(skb);
455 return ids->ids[hash - ids->n_ids * reciprocal_divide(hash, ids->rn_ids)];
459 * ovs_vport_receive - pass up received packet to the datapath for processing
461 * @vport: vport that received the packet
462 * @skb: skb that was received
463 * @tun_info: tunnel (if any) that carried packet
465 * Must be called with rcu_read_lock. The packet cannot be shared and
466 * skb->data should point to the Ethernet header. The caller must have already
467 * called compute_ip_summed() to initialize the checksumming fields.
469 void ovs_vport_receive(struct vport *vport, struct sk_buff *skb,
470 struct ovs_tunnel_info *tun_info)
472 struct pcpu_sw_netstats *stats;
473 struct sw_flow_key key;
476 stats = this_cpu_ptr(vport->percpu_stats);
477 u64_stats_update_begin(&stats->syncp);
479 stats->rx_bytes += skb->len;
480 u64_stats_update_end(&stats->syncp);
482 ovs_skb_init_inner_protocol(skb);
483 OVS_CB(skb)->input_vport = vport;
484 OVS_CB(skb)->egress_tun_info = NULL;
485 error = ovs_flow_key_extract(tun_info, skb, &key);
486 if (unlikely(error)) {
491 ovs_dp_process_packet(skb);
495 * ovs_vport_send - send a packet on a device
497 * @vport: vport on which to send the packet
500 * Sends the given packet and returns the length of data sent. Either ovs
501 * lock or rcu_read_lock must be held.
503 int ovs_vport_send(struct vport *vport, struct sk_buff *skb)
505 int sent = vport->ops->send(vport, skb);
507 if (likely(sent > 0)) {
508 struct pcpu_sw_netstats *stats;
510 stats = this_cpu_ptr(vport->percpu_stats);
512 u64_stats_update_begin(&stats->syncp);
514 stats->tx_bytes += sent;
515 u64_stats_update_end(&stats->syncp);
516 } else if (sent < 0) {
517 ovs_vport_record_error(vport, VPORT_E_TX_ERROR);
520 ovs_vport_record_error(vport, VPORT_E_TX_DROPPED);
526 * ovs_vport_record_error - indicate device error to generic stats layer
528 * @vport: vport that encountered the error
529 * @err_type: one of enum vport_err_type types to indicate the error type
531 * If using the vport generic stats layer indicate that an error of the given
534 static void ovs_vport_record_error(struct vport *vport,
535 enum vport_err_type err_type)
537 spin_lock(&vport->stats_lock);
540 case VPORT_E_RX_DROPPED:
541 vport->err_stats.rx_dropped++;
544 case VPORT_E_RX_ERROR:
545 vport->err_stats.rx_errors++;
548 case VPORT_E_TX_DROPPED:
549 vport->err_stats.tx_dropped++;
552 case VPORT_E_TX_ERROR:
553 vport->err_stats.tx_errors++;
557 spin_unlock(&vport->stats_lock);
560 static void free_vport_rcu(struct rcu_head *rcu)
562 struct vport *vport = container_of(rcu, struct vport, rcu);
564 ovs_vport_free(vport);
567 void ovs_vport_deferred_free(struct vport *vport)
572 call_rcu(&vport->rcu, free_vport_rcu);
575 int ovs_tunnel_get_egress_info(struct ovs_tunnel_info *egress_tun_info,
577 const struct ovs_tunnel_info *tun_info,
583 const struct ovs_key_ipv4_tunnel *tun_key;
587 if (unlikely(!tun_info))
590 tun_key = &tun_info->tunnel;
591 saddr = tun_key->ipv4_src;
592 /* Route lookup to get srouce IP address: saddr.
593 * The process may need to be changed if the corresponding process
594 * in vports ops changed.
607 /* Generate egress_tun_info based on tun_info,
608 * saddr, tp_src and tp_dst
610 __ovs_flow_tun_info_init(egress_tun_info,
611 saddr, tun_key->ipv4_dst,
618 tun_info->options_len);
623 int ovs_vport_get_egress_tun_info(struct vport *vport, struct sk_buff *skb,
624 struct ovs_tunnel_info *info)
626 /* get_egress_tun_info() is only implemented on tunnel ports. */
627 if (unlikely(!vport->ops->get_egress_tun_info))
630 return vport->ops->get_egress_tun_info(vport, skb, info);